Automorphisms of the loop and arc graph of an infinite-type surface Anschel Schaffer-Cohen December 17, 2019 Abstract We show that the extended based mapping class group of an infinite- type surface is naturally isomorphic to the automorphism group of the loop graph of that surface. Additionally, we show that the extended map- ping class group stabilizing a finite set of punctures is isomorphic to the arc graph relative to that finite set of punctures. This extends a known re- sult for sufficiently complex finite-type surfaces, and provides a new angle from which to study the mapping class groups of infinite-type surfaces. 1 Introduction and Main Result In the tradition of geometric group theory, mapping class groups are often stud- ied via their actions on metric spaces, most famously the Teichm¨uller space of the underlying surface. The mapping class group of a finite-type orientable sur- face also acts especially nicely on several simplicial complexes that have been defined for this purpose. These include the curve complex, the pants complex, and the arc complex. The curve complex and arc complex were defined for general infinite-type surfaces by Aramayona, Fossas, and Parlier in [AFP17], by analogy with existing definitions in the finite-type case and generalizing a definition for the plane minus a Cantor set given by Bavard in [Bav16]. Masur and Minsky [MM99] and Masur and Schleimer [MS13] showed that both these complexes are infinite- arXiv:1912.06774v1 [math.GT] 14 Dec 2019 diameter Gromov hyperbolic in the finite case, and Hensel, Przytycki, and Webb [HPW15] give an especially nice proof that the hyperbolicity constant does not depend on the surface. In addition, Ivanov [Iva97], Korkmaz [Kor99], Luo [Luo00], and Irmak and McCarthy [IM10] showed that the extended mapping class group is isomorphic to the automorphism groups of both complexes. Hern´andez Hern´andez, Morales, and Valdez recently showed in [HHMV18] that in the infinite-type case the mapping class group is still isomorphic to the automorphism group of the curve graph, which was also shown independently by Bavard, Dowdall, and Rafi in [BDR18]. However, the curve graph of an infinite-type surface has diameter 2, making it quasi-isometrically trivial. 1 On the other hand, the relative arc graph defined by Aramayona, Fossas, and Parlier is infinite-diameter Gromov hyperbolic, with a hyperbolicity con- stant that does not depend on the surface [AFP17]. Theorem 2 shows that, with appropriate restrictions, the mapping class group is precisely the group of automorphisms of this graph. We start with Theorem 1, which shows this result in the special case of the loop graph, an interesting structure in its own right. For our purposes, a surface Σ is a connected, oriented 2-manifold. We call Σ finite-type if π1pΣq is finitely generated, and infinite-type otherwise. Fix a basepoint p P Σ. A loop in Σ based at p is an unoriented simple closed curve in Σ starting and ending at p, considered up to isotopy fixing p. The loop graph LpΣ,pq of Σ is the graph whose vertex set is the set of all possible loops in Σ and where two vertices are connected by an edge if they have representatives intersecting only at p. The extended based mapping class group MCG˚pΣ,pq of Σ is the group of (possibly orientation-reversing) homeomorphisms Σ Ñ Σ that fix the basepoint p, considered up to isotopy preserving p. By construction, MCG˚pΣ,pq acts on LpΣ,pq, giving a homomorphism MCG˚pΣ,pq Ñ AutpLpΣ,pqq. The main result of this paper is that this is in fact an isomorphism. Theorem 1. Given an infinite-type surface Σ and a basepoint p P Σ, the map MCG˚pΣ,pq Ñ AutpLpΣ,pqq induced by the action is an isomorphism. Irmak and McCarthy [IM10] provide a successful program for proving this theorem in the finite-type case: 1. Fix a maximal set of disjoint loops, called a triangulation, and observe that its complementary regions are all triangles. 2. Show that some important local properties—three loops bounding a tri- angle, two triangles being adjacent, etc.—can be defined in terms of the loop graph and are therefore preserved by automorphisms of that graph. We use some of these results as Facts 6 and 7. 3. Use these local properties to construct a homeomorphism inducing a given transformation of our fixed triangulation. 4. Show that a homeomorphism which fixes a triangulation actually fixes the entire loop graph. We use this result as Fact 8. In our extension to the infinite-type case we have an advantage, a disadvan- tage, and a trick. The advantage is of course that we can depend on the existing result for finite-type surfaces. The disadvantage is that “triangulations” in the infinite-type setting can be much more exotic, as seen in Section 2. They will in fact have some complementary regions that are not actually triangles. The trick is to notice that Irmak and McCarthy’s proof is more general than the result requires: they start by fixing an arbitrary triangulation, but for their proof (and ours) it is sufficient to follow this program with any single triangulation. Thus we can construct a particularly useful triangulation for the specific purpose of building our homeomorphism. 2 In Section 5, we generalize Theorem 1 to the relative arc graph, which we define as follows: let P be a finite set of punctures (i.e. isolated planar ends) of an infinite-type surface Σ. An arc in Σ relative to P is an unoriented em- bedded line whose ends approach two (not necessarily distinct) punctures in Σ, considered up to isotopy.1 The relative arc graph ApΣ, P q of Σ relative to P is the graph whose vertex set is the set of all possible arcs relative to P in Σ and where two vertices are connected by an edge if they have disjoint representatives. The extended mapping class group MCG˚pΣ, P q of (possibly orientation-reversing) homeomorphisms stabilizing (but not necessarily fixing) the set P acts on ApΣ, P q. Then we can prove the following: Theorem 2. Given an infinite-type surface Σ and a finite set P of punctures of Σ, the map MCG˚pΣ, P q Ñ AutpApΣ, P qq induced by the action is an iso- morphism. There are two special cases to keep in mind: first, if P consists of a single puncture p, then ApΣ, P q“ LpΣYtpu,pq, so Theorem 2 really is a generalization of Theorem 1. Second, if Σ has finitely many punctures, then P can contain all of them and so MCG˚pΣ, P q“ MCG˚pΣq and we have a bona fide action of the mapping class group. The idea of the proof of Theorem 2 will be to reduce it to that of Theorem 1 by picking a puncture p P P and considering LpΣ Ytpu,pq as an induced subgraph of ApΣ, P q. The main hurdle will therefore be to prove that an auto- morphism of the arc graph preserves properties like “this arc is actually a loop” and “these two loops are based at the same puncture”. Acknowledgements The author would like to acknowledge the immense help of his adviser, David Futer, as well as fruitful conversations with Priyam Patel and Nicholas G. Vlamis and an insightful email from Edgar Bering. 2 Triangulations of infinite-type surfaces Irmak and McCarthy [IM10] define a triangulation as a maximal set of disjoint arcs, or equivalently as a set of arcs whose complementary regions are all tri- angles. Hatcher [Hat91], allowing for punctures not in P , observes that the complementary regions of a triangulation may also include punctured mono- gons. In the infinite-type case triangulations are considerably more exotic, as we shall see. So we simply define a triangulation2 to be a maximal clique in ApΣ, P q. The following facts are then immediate: Lemma 3. Given a set T of arcs and an automorphism f of ApΣ, P q, T is a triangulation if and only if fpT q is a triangulation. 1An isotopy of an arc cannot change its endpoints, since they are not included in Σ. 2This somewhat misleading name is (we hope) justified by the fact that it is a straightfor- ward extention of the usage in [IM10] and [Hat91]. 3 Lemma 4. Any set of disjoint arcs can be extended to a triangulation. Proof. This follows from Zorn’s Lemma. A finite-type surface admits countably many triangulations; each triangula- tion has the same number of loops depending only on the Euler characteristic of the surface; and any two triangulations are connected by a finite sequence of elementary moves, in which an arc α is removed from the triangulation and replaced with β, where the geometric intersection number ιpα, βq “ 1 [Hat91]. An infinite-type surface, on the other hand, admits uncountably many triangu- lations by application of Lemma 4, each with countably many arcs, and so most pairs of triangulations will not be connected by elementary moves. The topology of a triangulation is also interesting in the infinite-type setting: consider a finite union C of small circles centered at each p P P that intersects each arc finitely many times. By compactness, there must be points on C that are the limits of points on distinct arcs of the triangulation. A priori there is no reason to assume these limit points are contained in the triangulation at all, and so the triangulation may not be closed—in fact, we conjecture that it never is.